"In Clinical Trials" as a Lookup and One-hop Support Graph

NCATSTranslator / Feedback

A repo for tracking gaps in Translator data and finding ways to fill them.

7 stars 0 forks source link

"In Clinical Trials" as a Lookup and One-hop Support Graph #848

Closed sstemann closed 1 month ago

sstemann commented 3 months ago

It looks like we are using in clinical trials for both lookups and one-hop support graphs. I think that if there is a phase 3/4 clinical trials the correct predicate to use is Treats

But we're seeing it both as a lookup (in this case by BTE https://ui.test.transltr.io/main/results?l=Sepsis&i=HP:0100806&t=0&r=0&q=1357341d-e369-463e-bb6a-6c17f7c3908d)

and as a support graph (in this case by Improving Agent https://ui.test.transltr.io/main/results?l=Crohn%27s%20Colitis&i=MONDO:0005532&t=0&r=0&q=3642d7b7-2879-4e61-99b2-6bbc78949ea4)

neither have the CT links directly

sstemann commented 3 months ago

adding this spawned from #846

gglusman commented 3 months ago

The original definition for when to make a 'treats' assertion was for "a trial that successfully passed Phase 3, or a Phase 4 trial". But, the clinical trials database has no such notion as "successfully passed", so in the Clinical Trials KG I'm relying only on presence of Phase 4 to make 'treats' assertions.

sierra-moxon commented 2 months ago

@gglusman - do you imagine these results from BTE also be subsumed by CTKP? @andrewsu - does "Guppy" tag make sense here in light of the TACT discussion this morning?

gglusman commented 2 months ago

I am rather confused by the evidence being displayed. The lookup-edge claim is that cefazolin is in clinical trials for sepsis, and there's a 'treats' edge together with it. All the evidence displayed, though, is about product approvals from DailyMed, not a single one is from clinicaltrials (yes, I did scroll down the list):

This is a straight ingest from ChEMBL, so yes, this would be covered by CTKP. If there are phase 4 trials as suggested, there will be a 'treats' assertion already present in the KG.

sierra-moxon commented 1 month ago

Discussed CTKP in TACT today and there will be a push to harmonize CTKP ingests (vs. Chembl/others) in the next sprint.
@andrewsu - maybe the only change with that would be that we'd see the "in clinical trials" edge from CTKP as a support to a "treats" assertion from BTE. (I know we're planning on hearing from CQS folks next architecture meeting - another reason to move this to hammerhead).

andrewsu commented 1 month ago

Looks like the desired behavior is being observed (at least on test):

I think this satisfies the "ask" of BTE, but let us know if not!

gglusman commented 1 month ago

@andrewsu That 'in_clinical_trials_for' support edge comes from ChEMBL, not from CTKP. I'm not sure this answers Sierra's ask.

andrewsu commented 1 month ago

@gglusman I interpreted this issue as primarily about the structure of the treats and in_clinical_trials_for edges. But yes, you're right, the source of these edges (CTKP vs chembl) is also mentioned here. Is the Automat CTKP already available in the ITRB environment? It looks like the smartAPI annotation only shows a non-ITRB dev instance. If that's right, I suggest we close this issue and leave the data source swap for a different issue, synced with the next sprint.

gglusman commented 1 month ago

The Plover CTKP will be available in ITRB/CI next Friday.

andrewsu commented 1 month ago

Just so we can separate the two topics here, I'm going to close this issue that is primarily about the structure of results, and I created a new issue to track the swap of data sources for in_clinical_trials_for edges. Please let me know if that doesn't fully address the concerns!